Variable period adaptive genetic algorithm

Self-adapting genetic algorithm has two main factors contributing to its improved performance. The first is the effect of the progress of the evolution process where the fitness of the population improves as the number of generation increases. The second is the improvement due to the choice of the probabilities for the various genetic operators. In this paper, we propose a scheme that isolates the contributions of these two factors through the introduction of two competing populations. These two concurrent populations provide the necessary feedback to either prolong the duration of a good choice of the parameter setting or shorten that of a poor choice. Results from several numerical experiments have shown that the proposed scheme provides favorable performance over existing methods.